Tg

Pa

b

a

ARRA

KNLKEI

1

pamgtp(

oenStmSp

U

0h

Agriculture, Ecosystems and Environment 156 (2012) 49– 56

Contents lists available at SciVerse ScienceDirect

Agriculture, Ecosystems and Environment

jo ur n al homepage: www.elsev ier .com/ lo cate /agee

he effect of introducing a winter forage rotation on CO2 fluxes at a temperaterassland

aul Leahya,b,∗, Gerard Kielya,b

Department of Civil & Environmental Engineering, University College Cork, IrelandEnvironmental Research Institute, University College Cork, Ireland

r t i c l e i n f o

rticle history:eceived 16 February 2011eceived in revised form 16 April 2012ccepted 7 May 2012

eywords:et ecosystem exchangeand usealeddy covariance

a b s t r a c t

Temperate grasslands have the potential to sequester carbon, helping to mitigate rising atmosphericCO2 concentrations. The ability of grasslands to absorb CO2 is influenced by site elevation, soil type,management practices, climate and climatic variability. There is a need for long-term observations andfield experiments to quantify the effects of the key drivers of management and climate variability. Thispaper presents over 4 years of eddy covariance measurements of CO2 flux over a managed temperategrassland site in south-east Ireland. For the first 2 years the entire study area was under grass. Duringthe second 2 years a winter forage crop was grown over part of the site. The site was found to have anet uptake of CO2 during all years. However, the magnitude of the CO2 uptake varied considerably fromyear to year, with a maximum net uptake of 1.32 kg CO2 m−2 in 2004, a year with no winter forage crop.

reland Net uptakes were much lower in the 2 years of mixed grass and kale cultivation, but detailed analysis ofthe measurement footprint and statistical comparisons showed that this was not due to the introductionof the forage rotation. For a short period following sowing of the forage crop, daytime CO2 uptake wasless than that of the area under grass, but over subsequent months daytime CO2 uptake of the kale areasrecovered strongly and exceeded that of the grass areas. The net effect over the year following kaleplanting is close to CO2-neutral.

. Introduction

Grasslands cover 23% of the planet’s land area, with the tem-erate grassland and shrubland biome representing approximately

third of this (Lal, 2004). Janzen (2004) estimated that approxi-ately 9% of world net primary production occurs in temperate

rasslands and shrublands, and that these areas contain 15% ofhe global soil organic carbon pool. Within Ireland, grassland is therincipal land cover, occupying some 57% of the country’s land areaEaton et al., 2008).

Some established grasslands are believed to have a net uptakef atmospheric CO2 at annual timescales (Byrne et al., 2007; Jacobst al., 2007), but temperate, managed grasslands tend to have aet emission of CO2 over the winter months (Jaksic et al., 2006;kinner, 2007). A growing body of evidence suggests that long-erm carbon sequestration in grassland is largely controlled by

anagement practices (Soussana et al., 2004; Ammann et al., 2007;kinner, 2008; Wohlfahrt et al., 2008). It follows that it should beossible to increase the rate of sequestration by optimising the

∗ Corresponding author at: Department of Civil & Environmental Engineering,niversity College Cork, Ireland. Tel.: +353 21 4902017; fax: +353 21 4276648.

E-mail address: paul.leahy@ucc.ie (P. Leahy).

167-8809/$ – see front matter © 2012 Elsevier B.V. All rights reserved.ttp://dx.doi.org/10.1016/j.agee.2012.05.001

© 2012 Elsevier B.V. All rights reserved.

site management. Lal (2004) estimated that the annual potentialfor carbon (C) sequestration in the world’s pastures is almost aquarter of the combined annual C addition to the atmosphere dueto fossil fuel burning, cement production and land use change.

CO2 fluxes in pasture systems are influenced by managementactivities and by local climatic variability and it is often difficult todeconvolve these two influences in uncontrolled field observations.This is further complicated by the differing long-period variabilityand lagged responses to environmental factors of the different com-ponents of the CO2 flux, namely plant respiration, soil respirationand photosynthesis (Stoy et al., 2009). In a controlled experimenton an annual Mediterranean grassland, Chou et al. (2008) observedincreases in soil CO2 efflux following artificial wetting events andsuggested that the seasonal distribution of rainfall has an importantinfluence on ecosystem CO2 fluxes on the annual timescale. Lohilaet al. (2004) examined the effects of a spring barley rotation onCO2 exchange from a grassland on an improved peat soil in south-ern Finland and found a reduction in CO2 uptake during the periodof barley cultivation when compared to grass cultivation. Heavybiomass removal from a mature, temperate, North American grass-

land was found to render it a net C source, despite a small annualCO2 uptake (Skinner, 2008), whereas a similar, single-year studyfound a partially grazed, partially harvested grassland in southernIreland to be a small net C sink (Byrne et al., 2007).

5 stems and Environment 156 (2012) 49– 56

Ictwhygtcsfldontt

eot2iohtdpst

cniiAttffu

2

2

(d(m1tf

oMtesN

agi

−300 −200 −100 0 100−200

−100

0

100

200

flux tower

m

m

12

34

56

7

↑ N

2fluxes were measured using the eddy covariance technique. A fast,open-path infrared gas analyser (Li-7500, Li-Cor, USA) and an ultra-sonic anemometer (RM81000, R. M. Young, USA) were placed at

Table 1Dates of cutting and reseeding management events adjacent to the flux tower.

Date Day of year Paddock(s) Event

31/05/2004 152 5, 6, 7 Grass cut06/06/2005 157 1, 2, 3, 4, 5, 6, 7 Grass cut01/08/2005 213 1, 2, 3, 4, 5, 6, 7 Grass cut31/05/2006 151 1, 2, 3, 4, 5, 6, 7 Grass cut08/06/2006 159 5, 6 Kale sown03/08/2006 215 1, 2, 3, 4, 7 Grass cut18/04/2007 108 5, 6 Grass sown

0 P. Leahy, G. Kiely / Agriculture, Ecosy

In northwest European temperate grasslands such as those inreland, dairy cattle are usually housed indoors and fed on grass orereal silages with concentrate supplements over the low produc-ivity winter period (Hennessy et al., 2006). The practice of outdoorinter feeding on forage crop rotations, such as cereals or brassicas,as demonstrated agronomic benefits such as higher dry matterield and improved animal body condition compared to purelyrass-based winter feeding systems (Keogh et al., 2009). Despitehese advantages grass silage with supplements remains the typi-al source of winter cattle feed in Ireland, and there have been fewtudies of how winter forage rotations on grasslands affect the CO2uxes between such ecosystems and the atmosphere. There is evi-ence from a long-term experiment in Belgium that some functionsf the soil biota may be lower in ley-arable rotations than in perma-ent grasslands (van Eekeren et al., 2008), which leads to concernhat there may be a loss of soil organic carbon in such systems dueo repeated soil disturbance.

The eddy covariance (EC) technique has become well-stablished as a method of measuring surface–atmosphere fluxesf energy and trace gases, with several sites now observing con-inuously for 5 years or longer (Jacobs et al., 2007; Wohlfahrt et al.,008; Haslwanter et al., 2009). The technique is suitable for measur-

ng fluxes in the area adjacent to the sensor, but extended networksf flux observation sites together with satellite remote sensing haveelped to address the issue of quantifying CO2 exchange at the con-inental scale (Xiao et al., 2008). For comparison of fluxes fromifferent treatments, a moveable EC tower or ideally a system ofaired towers may be used (Davis et al., 2010). However, in thistudy we use a single EC tower combined with footprint analysiso separate fluxes originating from different areas.

The aim of this article is to examine whether a winter foragerop rotation at a temperate grassland increases or decreases theet annual CO2 uptake of the site, and to compare the difference,

f any, to the always-present, climate-driven interannual variabil-ty, in the context of a long-term CO2 flux monitoring campaign.

multi-year series of eddy covariance observations of net ecosys-em exchange (NEE) of CO2 at a managed agricultural site is usedo address these aims. Flux look-up tables based on environmentalactors and analysis of the eddy covariance footprint will be used toacilitate comparison between the areas under winter forage andnder grass.

. Methods

.1. Site

The study site is at Johnstown Castle Estate, Co. Wexford, Ireland52◦18′N; 6◦30′W). The measurement area is a flat grassland (pre-ominantly Lolium perenne) divided by wire fences into paddocksFig. 1). Brown earth is the dominant soil type within the measure-

ent area, with a mean organic carbon content of 3.9% in the top0 cm. A lightly-used road lies approximately 200 m west of theower behind a stone wall of 1 m height. The prevailing wind blowsrom the south-west.

Paddocks under grass were either (a) used for grazing through-ut the season on 3–4 week rotation or (b) harvested once in lateay or early June and used for grazing on 3–4 week rotation for

he rest of the grazing season, or (c) cut twice, once in late May orarly June, once in late July or early August and grazed late in theeason. The grazing season typically runs from mid-February to lateovember.

Most of the paddocks were grazed in 2004, with N inputs ofpproximately 240 kg ha−1, half of this applied as urea early in therowing season, and half as compound ammonium nitrate (CAN)n three applications between June and August. All paddocks were

Fig. 1. Site layout showing paddock boundaries and micrometeorological towerlocation, ringed with dotted circles of radius 100 m, 200 m and 300 m.

used for silage in 2005, with a total N application of 280 kg ha−1 asurea and CAN. Most of the site was used for silage in 2006 again, buta winter forage rotation was introduced to part of the site duringthis year. After the first grass harvest on June 8th, two paddocks (5and 6; Fig. 1) were limed and sown with kale (Brassica oleracea) forwinter forage and subsequently fertilised with 34 kg N ha−1 as CAN.The kale was grazed in situ between November 2006 and February2007. In 2006 the non-kale paddocks were used for silage, cut onMay 31st and again on August 3rd, and received similar fertiliserapplications to the previous year. Grazing did not commence untilmid-March in 2007, with paddocks 1 and 2 the first to be grazed. Thekale paddocks of 2006 (5 and 6) were ploughed, rotovated, rolledand reseeded with grass in April 2007. After the first grass harvestof 2007, in late May, paddocks 1 and 2 (Fig. 1) were harrowed andthen sown with kale to be used for in situ grazing over the winterof 2007/2008. The kale plots were fertilised in three applicationsbetween June and August 2007 with a total of 136 kg N ha−1. Alsoin 2007, two grass paddocks for grazing (3 and 4) were tilled to adepth of 30 cm, harrowed and reseeded with grass between August27th and 29th, having received a total of 205 kg N ha−1. No kalewas sown in 2008, and paddocks 1 and 2 were reseeded with grass,and paddocks 1–6 were grazed. The area beyond paddocks 1–7 to adistance of 400 m from the flux tower remained under grassland forthe entire duration of the study. The main harvesting and sowingevents within the measurement area are summarised in Table 1.

2.2. Measurements

Measurements of trace gas fluxes and meteorological quan-tities were made from January 2004 to September 2008. CO

23/05/2007 143 1, 2, 7 Grass cut05/06/2007 156 1, 2 Kale sown27/07/2007 208 7 Grass cut29/08/2007 241 3, 4 Grass resown

stems

2v1apThCgfsmtfplted

2

fire2t◦

tsiti

leyhfg(ec

•

•

•

•

•

P((

2

p

P. Leahy, G. Kiely / Agriculture, Ecosy

.5 m above the surface in order to measure the CO2 and waterapour concentrations and the three-dimensional wind speed at0 Hz frequency. Fluxes were computed at a 30 min interval by

datalogger (CR23X, Campbell Scientific, USA). Measurements ofhotosynthetically active radiation (PAR; PAR-lite, Kipp & Zonen,he Netherlands), soil temperature, air temperature and relativeumidity (HMP45C, Vaisala Inc., Finland), soil moisture (CS615,ampbell Scientific), long and short wavelength incoming andround-reflected radiation (CNR 100, Kipp & Zonen) and rain-all (ARG 100, Campbell Scientific) were also made at the studyite. Paddock boundaries were mapped using GPS surveying. Theicrometeorological sign convention is used for fluxes throughout

his paper, with a positive flux indicating transfer to the atmosphererom the surface. The site is part of a long-term flux monitoringroject on an operational farm, and management was therefore

argely outside the control of the flux measurement team. Thereforehe layout of the treatments was not optimal for conducting a sci-ntific experiment. Overcoming this limitation necessitated carefulata analysis.

.3. Data analysis and quality control

Raw 10 Hz wind, temperature, H2O and CO2 values were pre-ltered by the datalogger to remove extreme values outside theeliable range of the instruments or which were unrealistic in thisnvironment. Horizontal wind (u, v) speed values greater than5 m s−1 were excluded, as were vertical (w) speed values greaterhan 5 m s−1. Sonic temperature values outside the range [−10, 40]C; CO2 concentrations outside [5, 35] �mol m−3; H2O concentra-ions outside [50, 1050] mmol m−3 were also excluded. If any oneample of a quantity failed one of the above tests, the correspond-ng samples of all the others were also excluded. In addition, if morehan 25% values were excluded from any 30-min interval, the entirenterval was treated as a gap.

After pre-filtering, interval means and covariances were calcu-ated using custom routines written in the Matlab programmingnvironment (Mathworks, USA). Flux data were double rotated foraw and pitch correction (Wilczak et al., 2001). Temperatures andeat fluxes derived from ultrasonic measurements were corrected

or non-zero humidity (Massman and Lee, 2002). Measured 30 minas fluxes were also corrected by the addition of the WPL termWebb et al., 1980). Processed flux values were subjected to sev-ral quality control tests and were discarded if any of the followingriteria were met:

Flux values outside of the expected range for this ecosystem (e.g.uptake fluxes at night-time and values outside the prescribedseasonal ranges in Table 2) were ignored.Values corresponding to u∗ (frictional velocity) value of less than0.15 m s−1.Values where either the yaw or pitch rotation was greater than0.15 radian.Values calculated during intervals where the mean horizontalwind speed was greater than 25 m s−1, the coefficient of variationof CO2 or water vapour concentration was greater than 20%.Values from intervals where the quotient 〈u′w′〉/|u|2 was greaterthan 0.25.

rolonged gaps due to instrumental malfunctions occurred in 2004day numbers 231–237 and 270–302), 2005 (days 109–117), 2006days 324–336), 2007 (days 247–255) and 2008 (days 192–205).

.4. Gap filling procedure

Gaps in measured CO2 fluxes were filled using a look-up tablerocedure. Each complete year was divided into six periods, with

and Environment 156 (2012) 49– 56 51

the period boundaries being determined by site management deci-sions and phenology (Table 2). The first period of each year runsfrom the start of the year to the beginning of the growing season,usually indicated by a net daily uptake of CO2. The second periodruns from the start of the growing season to the first grass harvest.The third period runs from the first harvest through to the recov-ery of the ecosystem’s leaf area, as indicated by a net daily uptakeof CO2. The fourth period covers the period from leaf recovery tothe second harvest. The fifth period runs from the second harvestto the beginning of the autumnal downturn. This is defined as thetime when daily NEE turns from negative to positive (due to thereduction in gross primary productivity). The sixth and final periodcovers the low productivity period from autumn to the end of theyear, characterised by net emission fluxes as respiration dominatesthe NEE.

A two-dimensional look-up table was then derived from mea-sured data within each period. The two independent variables usedto derive the look-up tables were soil temperature measured at5 cm below the surface and above-canopy PAR. The set of availableflux measurements were classified into 64 subsets based on eightranges of PAR and eight ranges of soil temperature. The lowest PARrange was specified to run from zero to 50 �mol quanta m−2 s−1,in order to contain all fluxes measured under low-light conditionsor at nighttime. The boundaries of the remaining seven PAR binswere set such that each bin accounted for an equal mass of mea-sured CO2 flux. Soil temperature range boundaries were specifiedsuch that each bin contained the same mass of measured nighttimeflux.

A look-up table offers several advantages over regression-basedgap filling methods (e.g. (Novick et al., 2004; Jaksic et al., 2006)): itis purely data-driven and therefore requires no assumptions to bemade about the functioning of the ecosystem processes; it obviatesthe need for separate daytime and nighttime gap filling functions asthe PAR bin boundaries can be set to differentiate between daytimeand nighttime fluxes; and it can implicitly allow for net daytime res-piration, unlike some univariate nonlinear regressions, for examplea PAR-saturation curve such as a nonrectangular hyperbola (e.g.(Haxeltine and Prentice, 1996)) which can only produce uptakefluxes. The values in the tables corresponding to the lowest PARrange are effectively a nighttime look-up table in a single vari-able, namely soil temperature. The look-up table is a useful gapfilling method when the quantity of interest is fluxes measuredover timescales of longer than one day. It has the further advantageof allowing groupwise comparisons between flux measurementsfrom the same PAR, soil temperature bins across different years.The technique has been found to compare well with other gap fillingmethods over short and long timescales (Moffat et al., 2007).

In this study, the look-up table performed two functions. Firstly,it was used in the usual way to estimate fluxes from the controllingenvironmental parameters (e.g. PAR, Tsoil) during periods of missingor poor quality flux data. However, gap-filling with a look-up tableis not without disadvantages. There is a trade-off between the num-ber of PAR and soil temperature bins chosen and the number of goodflux measurements available in each bin to derive the table. There-fore, due to the limited number of bins, the small-scale variation ingap filled flux values within each bin is lost with this method. A fur-ther disadvantage is that if fixed boundaries are prescribed for thebins of the independent variables, there is no guarantee that everybin will be populated by good flux measurements, as the indepen-dent variables may not be evenly distributed across the bins. Thisposes the problem of selecting a suitable flux value to assign tothese empty bins, should a corresponding gap in the flux record

occur. For the purposes of calculating the annual flux totals over theentire measurement area, including both land uses, when a look-uptable bin was unpopulated the 5-day average of the flux value forthat time of day was used to fill the gap. The reader is referred to

52 P. Leahy, G. Kiely / Agriculture, Ecosystems and Environment 156 (2012) 49– 56

Table 2(a) Start and end day numbers defining periods for allowable seasonal flux magnitude ranges and for generating gap filling look-up tables for each year; (b) acceptance rangesin �mol m−2 s−1 for daytime and nighttime CO2 fluxes in each period.

Period

1 2 3 4 5 6

(a)2004 1–75 76–120 121–152 153–210 211–280 281–3652005 1–100 101–156 157–172 173–213 214–280 281–3652006 1–100 101–151 152–172 173–215 216–280 281–3652007 1–100 101–156 157–190 191–240 241–310 311–3652008 1–90 91–158 159–211 212–246 – –

, 15]

5]

tesite

2

u(acFswaicacnsmeg

ltiokMwuawd(makpd

d

(

the study period, with moisture content well above wilting point(ca. 12 % of total volume) throughout. For the single-crop years of2004, 2005 and 2006, the largest net annual CO2 uptake occurredin 2004 and the lowest in 2006 (Table 3).

Table 3Annual sums of gap filled net ecosystem exchange of CO2 over the entire measure-ment area with associated uncertainty estimates, percentage of the annual recordthat was gap-filled, and number of observations originating from each vegetationcover for the complete measurement years 2004–2007.

Year Ng Nk Gap-filled (%) NEE (g CO2 m−2)

(b)Day [−15, 5] [−25, 10] [−35Night [0, 7] [0, 10] [0, 1

he studies of Falge et al. (2001), Reichstein et al. (2005) and Moffatt al. (2007) for a full discussion of flux gap filling techniques. Theecond use of the look-up table was to enable like-for like compar-sons of fluxes emanating from the two different land uses withinhe same soil temperature and PAR bins. In this way, managementffects can largely be separated from effects of climatic variability.

.5. Comparing fluxes of different land uses

The approximate analytical model of Hsieh et al. (2000) wassed to estimate the areas contributing to the observed CO2 fluxFig. 2). The model produces a vector of cumulative source strengths a function of distance from the sensor. This allows the dominantrop cover associated with each measurement to be determined.or each 30-min flux interval, the footprint model was run and theource strength vector was used in combination with the measuredind direction to calculate the proportion of the flux originating in

reas under grass and under kale. The source strength vector wasntegrated along its intersection with the polygon enclosing therop type area. If one crop accounted for 66% or more of the flux for

given interval, then that interval was assigned to that particularrop type. Any intervals during which neither crop type was domi-ant (i.e. neither land use area accounting for more than 66% of theource flux) were ignored in subsequent analyses. No attempt wasade to account for the effects of inhomogenous surface cover, for

xample the week when the footprint was partially covered by cutrass and partially bare, immediately after kale sowing in 2007.

This resulted in two time series of data: one for the grass-and target area and one for the kale target area. Both series werehen binned based on PAR and soil temperature bins as describedn Section 2.4, within the time periods specified in Table 2. Fluxbservations in corresponding PAR, soil temperature bins in theale and non-kale populations were compared groupwise using theann–Whitney U-test from the Matlab statistics toolbox (Math-orks, USA). This test was chosen as it does not require the samplesnder comparison to be normally distributed when sample sizesre small (Devore, 1995). The null hypothesis in each comparisonas that the observations in corresponding grass and kale bins wererawn from the same population, and the chosen significance levelP) for rejection was 0.05. The method allows confounding environ-

ental factors (PAR and soil temperature) to be largely discountednd a direct comparison of CO2 fluxes from areas under grass andale to be made. The observations in each series were also com-ared in aggregate, i.e. over all bins within a period, in order toetermine if the two complete sets of observations differed.

In summary, the approach taken to comparing the flux from theifferent land use areas was:

1) Divide the flux time series into sub-annual time periods, asspecified in Table 2.

[−35, 15] [−25, 10] [−15, 5][0, 15] [0, 10] [0, 7]

(2) Classify each flux observation based on the instantaneous fetchand wind direction: series G (where grass land cover domi-nates) and series K (where kale dominates).

(3) Build a look-up table based on PAR and soil temperature (asdescribed above) for each series within each period.

(4) Statistically compare the average fluxes in like-for-like bins ofthe look-up tables for series G and series K.

An annual flux estimate from each land use type for the year2007/2008 was also calculated by separately gap filling series Gand series K over the entire year. For each individual 30-min fluxmeasurement, the associated uncertainty was estimated by thecorresponding standard deviation within the relevant bin in thelook-up tables. Uncertainty estimates for the full year were thenobtained by combining the 30-min measurement uncertainties inquadrature.

3. Results

The mean ratio of the modelled fetch distance (distance from thesensor incorporating 90% of flux) to sensor height ranged between273 (2008) and 295 (2006). There was considerable interannualvariability in the environmental parameters (PAR, soil temperature,rainfall and soil moisture; Fig. 3) and in monthly CO2 fluxes (Fig. 4).Summer 2004 was dry, with soil water filled pore space (WFPS)content remaining below 60% for much of May, June and July. Both2005 and 2006 were characterised by heavy spring and early sum-mer precipitation, with high soil moisture levels persisting untilearly June of each of the 2 years. After a relatively dry April, thesummer of 2007 was extremely wet compared to other years, esti-mated as the wettest in at least 9 years by Met Éireann, with highamounts of precipitation received between May and early August,and soil WFPS remained above 60% during the same period. Thelarge variations in March and April rainfall across all years did nottranslate into large variations in soil moisture as most rainfall runsoff at this time of year. Moisture stress was not encountered during

2004 8167 0 63 −1321 ± 2202005 10,376 0 53 −1261 ± 2832006 10,361 0 50 −658 ± 2082007 11,605 1490 47 −75 ± 188

P. Leahy, G. Kiely / Agriculture, Ecosystems and Environment 156 (2012) 49– 56 53

F (b) il( an wi

m(aowItoabt(

ggr

ig. 2. (a) Windrose for 2007 (number of 30-min intervals versus sector angle) andm) and wind direction for 2007. The dots represent the peak fetch distance and me

In 2006, the first year when kale was sown, none of the fluxeasurements were found to be dominated by the kale area

Table 3). This was due to the location of the kale paddocks rel-tive to the flux tower, with the nearest boundary being locatedutside the prevailing wind sector, approximately 100 m north-est of the tower, and separated from it by an area of grassland.

n 2007, when kale was sown in two paddocks adjacent to theower, lying mainly in the southwest to southeast quadrant, 1490f 17,520 half hourly flux measurements were from the kalerea, 11,605 from the grass area and the remainder could note ascribed to a dominant land use type, either because nei-her area dominated, or due to missing or poor quality dataTable 3).

Mann–Whitney U tests of sets of flux measurements (aggre-

ated over all bins of the look-up tables) over kale areas versusrass areas for periods three to six of 2007 (cf. Table 2) did notesult in a rejection of the null hypothesis, that the two sets of

1 2 3 4 5 0

500

1000

1500

Tot

al P

AR

[mol

qua

nta

m−

2 ]

2004

2 4 0

5

10

15

20

Tso

il

[° C]

2 4 0

50

100

150

200

Rai

nfal

l[m

m]

30−year average

j f m a m 0

50

100

WF

PS

[%]

Fig. 3. Monthly total PAR, mean soil temperature at 5 cm depth, total rainfall and mea

lustration of flux partitioning by land use type from calculated peak fetch distancend direction corresponding to each 30-min flux measurement.

fluxes were drawn from the same distribution. However, when theobserved fluxes within corresponding Tsoil and PAR bins were com-pared across grass and kale series, a majority of the paired binsexhibited a statistically significant difference in periods four andfive (Table 4). Therefore flux observations over kale and grass dur-ing these periods, controlled for environmental factors, are likelyto be drawn from different distributions.

If the two series of flux observations are separately gap filled andcumulatively summed, the cumulative difference in NEE betweeneach area can be compared (Fig. 5). For the first 40–50 days afterkale was sown, the grass areas had a higher CO2 uptake than thekale areas, however this trend reversed for the following 30–40days, and both areas had similar NEE for the remainder of the year.However, it should be noted that the period from day 247 to 255

was 100% gap filled due to instrumental downtime. At the end ofthe year, the annual NEE is −500 g CO2 m−2 for the grass area and−398 g CO2 m−2 for the kale area, therefore the difference in NEE is

6 7 8 9 10 11 12

’05 ’06 ’07 ’08

6 8 10 12

6 8 10 12

j j a s o n d

month

n soil water filled pore space (WFPS) at 5 cm depth, January 2004–August 2008.

54 P. Leahy, G. Kiely / Agriculture, Ecosystems and Environment 156 (2012) 49– 56

j f m a m j j a s o n d−500

0

5002004

j f m a m j j a s o n d−500

0

5002005

NE

E [g

CO

2 m−

2 s−

1 ]

j f m a m j j a s o n d−500

0

5002006

j f m a m j j a s o n d−500

0

5002007

Month

j f m a m j j a s o n d−500

0

5002008

Month

Fig. 4. Monthly gap filled net ecosystem exchange totals.

Table 4Mean daytime and nighttime non-gapfilled CO2 fluxes (�mol CO2 m−2 s−1) over grass- and kale-dominated areas during periods of 2007; percentage of non-empty grass andkale flux bins which differed at the 95% significance level. The period boundaries are listed in Table 2.

Period Mean flux (grass) Mean flux (kale) Significance of difference

Daytime Nighttime Daytime Nighttime % bins significantly different

Period 3 −3.8 4.7 −0.6 4.3 36−4.1−5.8−3.2

co

4

u

Fdba

Period 4 −1.5 3.7

Period 5 −0.5 2.6

Period 6 −2.9 1.5

a. 100 g CO2 m−2, less than the overall uncertainty in the differencef annual flux sums (± 338 g CO2 m−2).

. Discussion

The years 2004 and 2005 showed strong annual CO2 uptake,nder 100% grass cover. In early August 2005 a second silage cut

2007:156190 240 310 365 2008:90−300

−200

−100

0

100

200

300

400

500

Day of year, 2007/2008

NE

Eg −

NE

Ek [g

CO

2 m−

2 ]

ig. 5. Difference between NEE measured over grass (NEEg) and kale (NEEk) pad-ocks for 1 year from planting in 2007 (day number 156). Gap filling seasonoundaries are illustrated by the dashed vertical lines; the grey portion represents

period of 100% gap filling during 2007 due to an instrumental malfunction.

4.9 58 3.2 68 2.3 22

was taken from most of the study area, but monthly NEE duringthe subsequent months was similar to the corresponding monthsof 2004, when there was only a single cut in May. Annual rainfalltotals were close to the long-term average for both 2004 and 2005(Met Éireann, 2004–2008). Lower annual uptakes were recordedin the two subsequent years under mixed usage (2006 and 2007)and part of 2008. The extremely wet spring of 2006, with 131%of normal rainfall (Met Éireann, 2004–2008), delayed the onset ofpeak biomass accumulation, which was compounded by lower thanaverage PAR levels during May (Fig. 3). June CO2 fluxes were vari-able between 2004 and 2008, probably mainly due to variation inharvesting dates (Fig. 4 and Table 1). There was a net emission ofCO2 flux during the extremely wet June of 2007, when 223% of thenormal monthly rainfall was received (Met Éireann, 2004–2008);however this is likely to be at least partially due to the reseedingof part of the measurement area with winter kale. PAR was alsobelow average in June of 2007. June 2006 saw net emissions of CO2under different climatic conditions to 2007 (drier soils) but sim-ilar management conditions. In 2007 and 2008, the wet summerconditions may have suppressed biomass recovery after the firstsilage cut. CO2 uptake in July 2007 was less than in previous yearsand for most of the remainder of 2007 there were net monthlyemissions, possibly due to the extremely wet conditions followingthe 2007 harvest. Conditions during 2008 were close to normal, butJune was extremely wet, with 225% of average rainfall (Met Éireann,2004–2008).

The number of kale-dominated measurement intervals in 2007was a fraction of the number of grass-dominated measurements,

hence caution is necessary when attempting to compare the resultsfrom each area. Nevertheless, statistical tests of difference showsome evidence for a difference between the two areas in periodsthree through five of 2007 (approximately June through October;

stems

TasthrhrflstesawplUesust(aCpsnbd

siogioatewatd

cfmcaksoa2

scbbsccci

P. Leahy, G. Kiely / Agriculture, Ecosy

able 4), when many of the measured fluxes in corresponding kalend grass bins differ significantly. Immediately after the kale wasown on June 5th (day 156), the kale areas emitted more CO2 thanhe grass areas (Fig. 5). This is expected as the grass areas, althougharvested only 1 week before the kale was sown, would haveetained some above-ground biomass, and the kale areas wouldave been subjected to below-ground disturbance, stimulating soilespiration. However, 2 months after kale planting, daytime uptakeuxes appear to be stronger in the kale area, but nighttime emis-ions are also larger. The cumulative difference in NEE betweenhe two cultivation areas diminished as the comparison period wasxtended into the winter months (Fig. 5). This pattern bears someimilarities to a study of net ecosystem exchange of spring barleynd perennial grass on a peatland in Finland, where higher uptakesere observed from the barley crop than from the grass during theeak growing season, but the grass exhibited CO2 uptake over a

onger period due to its longer growing season (Lohila et al., 2004).nlike our study, the annual NEE of the barley cropped area greatlyxceeded that of the grass area. A comparison of three managementystems: no-till barley forage; conventional till barley forage; andndisturbed perennial grass–alfalfa mixture, also found elevatedoil CO2 fluxes in the conventionally tilled forage plot comparedo undisturbed perennial forage for several months after tillingJabro et al., 2008). The measurements of Willems et al. (2011) at andjacent site, showed that ploughed areas actually exhibited lowerO2 emissions than undisturbed areas, apart from a brief emissioneak lasting for less than 1 day after ploughing. This implies thatoil disturbance plays a minor role and that biomass is the domi-ant influence on overall fluxes, thus suggesting that the differenceetween grass and kale fluxes in our study is primarily due to theifference in crop cover.

Although a statistically significant difference has been demon-trated between the corresponding bins of flux observations, this intself does not conclusively prove that the land cover is the main, ornly, cause of the difference. The tilling and reseeding of two of therass paddocks in August 2007 introduced a further disturbancento the system and may contribute to the large net CO2 emissionf that month and the low mean daytime uptake fluxes over grassreas in period four of 2007 (Table 4). It has been shown elsewherehat the practice of adding lime affects soil CO2 emissions (Biasit al., 2008), but in this study only the paddocks under kale in 2006ere limed, therefore the 2007 grassland/kale comparison was not

ffected by liming. Heterogeneity of soil properties is another fac-or not considered here which could potentially contribute to aifference in NEE between the two areas.

Despite some evidence from the evolution of the difference inumulative NEE recorded over grass and kale areas, the overall dif-erence is small 1 year after kale sowing (Fig. 5) and lies within the

easurement uncertainty. The bias introduced by the different per-entage of gap filling for each subseries may be an influence here,nd the possibility also exists that delayed effects of the previousale rotation reduced productivity during these years, although ithould be noted that the paddocks under kale in 2006 occupiednly a small proportion of the area sampled by the EC system, sony such delayed effect would have been minimal in the case of007.

As EC measurements form part of a time series rather than aet of completely independent observations, the use of statisticalomparisons should be approached with caution, however it haseen attempted to control for variability in environmental factorsy means of partitioning flux observations into bins of PAR andoil temperature. It should be noted that soil temperature is not a

ompletely independent variable as the sowing of the winter foragerop affects the soil temperature through the removal of vegetationover. Furthermore, the effects of soil moisture are not explicitlyncluded in the look-up tables. The gap filling procedure exhibited

and Environment 156 (2012) 49– 56 55

some systemic bias, and tended to underestimate high emissionfluxes, implying that the annual fluxes presented in Table 3 may beunderestimated. The slopes of linear regressions of annual sets of30-min fluxes modelled using the look-up tables to contemporane-ous measured fluxes were in the range 0.7–0.8. Due to the chosenapproach of separately gap filling flux series measured over grassand kale, the difference in proportions of the flux data requiringgap-filling between grassland years (2004–2005; 2008) and themixed grass/kale cultivation years (2006 and 2007), is a possiblefurther source of bias in the final, gap-filled figures.

5. Conclusion

CO2 fluxes from an area under year round grassland cultivationand an adjacent, similar area under a kale winter forage rotationhave been measured and compared. Interannual climatic variabilityand its impact on the timing of grass harvesting introduce varia-tions in the magnitude, and sometimes even the sign, of grasslandNEE between the same months of different years. The largest differ-ences in net annual fluxes were between years with different cropcover. In general, annual uptakes were much larger in the 2 years ofgrass cultivation than in the years of mixed grass/kale cultivation.

The effect of climatic variability on the flux observations was iso-lated by binning flux observations into classes of soil temperatureand PAR and deriving flux look-up tables based on these bins to cal-culate annual flux sums. Annual CO2 uptakes ranged from 75 g CO2m−2 to 1321 g CO2 m−2. Although there is evidence for increasedrespiration and reduced CO2 uptake in the first weeks after kalesowing in this managed temperate grassland site, the CO2 uptakeof the kale area recovers strongly in the months after sowing, andsubsequently exceeds that of the grass area. Overall, this agronom-ically desireable winter forage rotation does not result in increasedCO2 emissions at least in the first year following sowing. It is recom-mended that further work be carried out in a controlled, replicatedfield experiment in order to quantify a complete C budget for eachmanagement practice.

Acknowledgements

The authors wish to thank Adrian Birkby, Kilian Murphy andViacheslav Voronovich of UCC and Deirdre Fay, Owen Carton, DonalDoyle and Gary Lanigan of Teagasc for assistance. This work wasfunded by the Irish Environmental Protection Agency’s STRIVEprogramme (project Celticflux; 2001-CD-C2-M1) and the EU 6thFramework project CarboEurope-IP. Paul Leahy also acknowledgesthe support of the Environmental Protection Agency and the StokesLectureship programme of Science Foundation Ireland.

References

Ammann, C., Flechard, C.R., Leifeld, J., Neftel, A., Fuhrer, J., 2007. The carbon budgetof newly established temperate grassland depends on management intensity.Agric. Ecosyst. Environ. 121 (June (1–2)), 5–20.

Biasi, C., Lind, S.E., Pekkarinen, N.M., Huttunen, J.T., Shurpali, N.J., Hyvonen, N.P.,Repo, M.E., Martikainen, P.J., 2008. Direct experimental evidence for the contri-bution of lime to CO2 release from managed peat soil. Soil Biol. Biochem. 40 (10),2660–2669.

Byrne, K.A., Kiely, G., Leahy, P., 2007. Carbon sequestration determined using farmscale carbon balance and eddy covariance. Agric. Ecosyst. Environ. 121 (4),357–364.

Chou, W.W., Silver, W.L., Jackson, R.D., Thompson, A.W., Allen-Diaz, B., 2008. The sen-sitivity of annual grassland carbon cycling to the quantity and timing of rainfall.Global Change Biol. 14 (6), 1382–1394.

Davis, P.A., Brown, J.C., Saunders, M., Lanigan, G., Wright, E., Fortune, T., Burke,J., Connolly, J., Jones, M.B., Osborne, B., 2010. Assessing the effects of agricul-

tural management practices on carbon fluxes: spatial variation and the need forreplicated estimates of net ecosystem exchange. Agric. For. Meteorol. 150 (4),564–574.

Devore, J.L., 1995. Probability and Statistics for Engineering and the Sciences, 4th ed.Duxbury Press, Pacific Grove, CA, USA.

5 stems

E

F

H

H

H

H

J

J

J

J

K

L

L

M

M

MMatamala, R., McNulty, S., Meyers, T.P., Monson, R.K., Munger, J.W., Oechel,

6 P. Leahy, G. Kiely / Agriculture, Ecosy

aton, J.M., McGoff, N.M., Byrne, K.A., Leahy, P., Kiely, G., 2008. Land cover changeand soil organic carbon stocks in the Republic of Ireland 1851–2000. ClimaticChange 91 (3–4), 317–334.

alge, E., Baldocchi, D., Olson, R., Anthoni, P., Aubinet, M., Bernhofer, C., Burba,G., Ceulemans, R., Clement, R., Dolman, H., Granier, A., Gross, P., Grunwald, T.,Hollinger, D., Jensen, N.-O., Katul, G., Keronen, P., Kowalski, A., Lai, C.T., Law, B.E.,Meyers, T., Moncrieff, J., Moors, E., Munger, J.W., Pilegaard, K., Rannik, U., Reb-mann, C., Suyker, A., Tenhunen, J., Tu, K., Verma, S., Vesala, T., Wilson, K., Wofsy,S., 2001. Gap filling strategies for defensible annual sums of net ecosystemexchange. Agric. For. Meteorol. 107 (1), 43–69.

aslwanter, A., Hammerle, A., Wohlfahrt, G., 2009. Open-path vs. closed-path eddycovariance measurements of the net ecosystem carbon dioxide and watervapour exchange: a long-term perspective. Agric. For. Meteorol. 149 (February(2)), 291–302.

axeltine, A., Prentice, I.C., 1996. A general model for the light-use efficiency ofprimary production. Funct. Ecol. 10 (October (5)), 551–561.

ennessy, D., O’Donovan, M., French, P., Laidlaw, A.S., 2006. Effects of date of autumnclosing and timing of winter grazing on herbage production in winter and spring.Grass Forage Sci. 61 (4), 363–374.

sieh, C.I., Katul, G.G., Chi, T.W., 2000. An approximate analytical model for footprintestimation of scalar fluxes in thermally stratified atmospheric flows. Adv. WaterResour. 23 (7), 765–772.

abro, J.D., Sainju, U., Stevens, W.B., Evans, R.G., 2008. Carbon dioxide flux as affectedby tillage and irrigation in soil converted from perennial forages to annual crops.J. Environ. Manage. 88, 1478–1484.

acobs, A.F.G., Heusinkveld, B.G., Holtslag, A.A.M., 2007. Seasonal and interannualvariability of carbon dioxide and water balances of a grassland. Climatic Change82 (May (1–2)), 163–177.

aksic, V., Kiely, G., Albertson, J., Oren, R., Katul, G., Leahy, P., Byrne, K., 2006. Netecosystem exchange of grassland in contrasting wet and dry years. Agric. For.Meteorol. 139 (3–4), 323–334.

anzen, H.H., 2004. Carbon cycling in earth systems – a soil science perspective. Agric.Ecosyst. Environ. 104 (3), 399–417.

eogh, B., French, P., McGrath, T., Storey, T., Mulligan, F.J., 2009. Comparison of theperformance of dairy cows offered kale, Swedes and perennial ryegrass herbagein situ and perennial ryegrass silage fed indoors in late pregnancy during winterin Ireland. Grass Forage Sci. 64 (1), 49–56.

al, R., 2004. Soil carbon sequestration to mitigate climate change. Geoderma 123(1–2), 1–22.

ohila, A., Aurela, M., Tuovinen, J.-P., Laurila, T., 2004. Annual CO2 exchange of apeat field growing spring barley or perennial forage grass. J. Geophys. Res. 109(D18116).

assman, W.J., Lee, X., 2002. Eddy covariance flux corrections and uncertainties inlong-term studies of carbon and energy exchanges. Agric. For. Meteorol. 113(1–4), 121–144.

et Éireann, 2004–2008. Weather Reports. Tech. rep., Met Éireann, Glasnevin,Dublin 9, Ireland. http://www.met.ie/climate/monthly-summary-archive.asp.

offat, A.M., Papale, D., Reichstein, M., Hollinger, D.Y., Richardson, A.D., Barr,A.G., Beckstein, C., Braswell, B.H., Churkina, G., Desai, A.R., Falge, E., Gove, J.H.,Heimann, M., Hui, D., Jarvis, A.J., Kattge, J., Noormets, A., Stauch, V.J., 2007. Com-prehensive comparison of gap-filling techniques for eddy covariance net carbonfluxes. Agric. For. Meteorol. 147 (3–4), 209–232.

and Environment 156 (2012) 49– 56

Novick, K.A., Stoy, P.C., Katul, G.G., Ellsworth, D.S., Siquira, M.B.S., Juang, J., Oren, R.,2004. Carbon dioxide and water vapor exchange in a warm temperate grassland.Oecologia 138 (2), 259–274.

Reichstein, M., Falge, E., Baldocchi, D., Papale, D., Aubinet, M., Berbigier, P., Bern-hofer, C., Buchmann, N., Gilmanov, T., Granier, A., Grunwald, T., Havrankova, K.,Ilvesniemi, H., Janous, D., Knohl, A., Laurila, T., Lohila, A., Loustau, D., Matteucci,G., Meyers, T., Miglietta, F., Ourcival, J.-M., Pumpanen, J., Rambal, S., Rotenberg,E., Sanz, M., Tenhunen, J., Seufert, G., Vaccari, F., Vesala, T., Yakir, D., Valentini,R., 2005. On the separation of net ecosystem exchange into assimilation andecosystem respiration: review and improved algorithm. Global Change Biol. 11(9), 1424–1439.

Skinner, H., 2007. Winter carbon dioxide fluxes in humid-temperate pastures. Agric.For. Meteorol. 144 (1–2), 32–43.

Skinner, R.H., 2008. High biomass removal limits carbon sequestration poten-tial of mature temperate pastures. 4th USDA Greenhouse Gas Symposiumon Positioning Agriculture and Forestry to Meet the Challenges of Cli-mate Change, Baltimore, MD, Feb. 6–8, 2007. J. Environ. Qual. 37 (4),1319–1326.

Soussana, J.-F., Loiseau, P., Vuichard, N., Ceschia, E., Balesdent, J., Chevallier, T.,Arrouays, D., 2004. Carbon cycling and sequestration opportunities in temperategrasslands. Soil Use Manage. 20 (2), 219–230.

Stoy, P.C., Richardson, A.D., Baldocchi, D.D., Katul, G.G., Stanovick, J., Mahecha,M.D., Reichstein, M., Detto, M., Law, B.E., Wohlfahrt, G., Arriga, N., Campos, J.,McCaughey, J.H., Montagnani, L., Paw U, K.T., Sevanto, S., Williams, M., 2009.Biosphere–atmosphere exchange of CO2 in relation to climate: a cross-biomeanalysis across multiple time scales. Biogeosciences 6, 2297–2312.

van Eekeren, N., Bommele, L., Bloem, J., Schouten, T., Rutgers, M., de Goede, R., Reheul,D., Brussaard, L., 2008. Soil biological quality after 36 years of ley-arable crop-ping, permanent grassland and permanent arable cropping. Appl. Soil Ecol. 40(3), 432–446.

Webb, E.K., Pearman, G.I., Leuning, R., 1980. Correction of flux measurements fordensity effects due to heat and water-vapor transfer. Quart. J. Roy. Meteorol.Soc. 106, 85–100.

Wilczak, J.M., Oncley, S.P., Stage, S.A., 2001. Sonic anemometer tilt correction algo-rithms. Bound.-Layer Meteorol. 99 (1), 127–150.

Willems, A.B., Augustenborg, C.A., Hepp, S., Lanigan, G., Hochstrasser, T., Kammann,C., Mueller, C., 2011. Carbon dioxide emissions from spring ploughing of grass-land in Ireland. Agric. Ecosyst. Environ. 144 (1), 347–351.

Wohlfahrt, G., Hammerle, A., Haslwanter, A., Bahn, M., Tappeiner, U., Cernusca, A.,2008. Seasonal and inter-annual variability of the net ecosystem CO2 exchangeof a temperate mountain grassland: effects of weather and management. J. Geo-phys. Res. 113 (D08110).

Xiao, J., Zhuang, Q., Baldocchi, D.D., Law, B.E., Richardson, A.D., Chen, J., Oren, R.,Starr, G., Noormets, A., Ma, S., Verma, S.B., Wharton, S., Wofsy, S.C., Bolstad,P.V., Burns, S.P., Cook, D.R., Curtis, P.S., Drake, B.G., Falk, M., Fischer, M.L., Fos-ter, D.R., Gu, L., Hadley, J.L., Hollinger, D.Y., Katul, G.G., Litvak, M., Martin, T.A.,

W.C., U, K.T.P., Schmid, H.P., Scott, R.L., Sun, G., Suyker, A.E., Torn, M.S., 2008.Estimation of net ecosystem carbon exchange for the conterminous UnitedStates by combining MODIS and AmeriFlux data. Agric. For. Meteorol. 148 (11),1827–1847.